Minimum Variance Control over a Gaussian Communication Channel

Professor James Freudenberg

Department of Electrical and Computer Engineering

University of Michigan

Abstract:

Much work has been devoted to the problem of finding the minimum channel capacity required to stabilize an unstable plant. Much less is known about the problem of achieving performance goals with a communication channel in the feedback loop. In this talk, we consider the problem of minimizing the variance of the plant output using a measurement obtained from an additive white Gaussian noise channel. There are two differences between this problem and the standard LQG problem: the presence of a power constraint at the channel input, and the ability to add compensation both before and after the noisy channel (the encoder and decoder). We use ideas from stochastic control, estimation, and information theory to design communication and control strategies to minimize a measure of the disturbance response variance.

Friday, February 8, 2008

3:30 – 4:30 p.m.

Rm. 1500 EECS